[linux-cirrus] Re: new Cirrus Logic Linux release
- From: Lennert Buytenhek <buytenh@xxxxxxxxxxxxxx>
- To: linux-cirrus@xxxxxxxxxxxxx
- Date: Sat, 2 Jun 2007 03:25:31 +0200
Someone forwarded this email to me. There are some things that I
would like to set straight.
On Fri, 01 Jun 2007 04:24:33 -0500, Brian Austin wrote:
> What does "code quality" give you?
>
> The issue I have with "mainstream" is that the code changes that would
> have to be done, do not actually make the chip and kernel any better.
> The performance of the maintream kernel vs. Cirrus kernel are really
> that different?
The upstream ep93xx ethernet driver was over 100% faster on very
common workloads last time I checked.
I did write a polite email to the Cirrus engineer that wrote the Cirrus
ethernet driver explaining the performance differences and what they
should change in their version of the driver to get the same performance
levels, so maybe the performance bugs in the Cirrus driver have been
fixed in the meanwhile, I don't know.
Anyway, not really relevant to the discussion.
> Anyway, It's not that I didnt want to work with Lennert, I did, I
> sent him free HW and pushed my bosses very hard to work something
> out, but then the communication broke down from a professional
> manner to flaming and bad mouthing and posting of private emails
> on public forums and that really pissed me off, and I dont have
> time to deal with it.
Let's set some things straight.
All of the initial upstream ep93xx work was done by me without Cirrus
involvement. After a while, Cirrus approached me (not the other way
round), telling me that they really appreciated my work and were
almost certainly going to pay me for it.
Then a couple months later, Cirrus suddenly changed their attitude and
more or less told me "We're not not going to pay you after all, and
we're not going to work with you either." This came totally out of
the blue for me, and wasn't because of anything I did, contrary to
what Brian is suggesting above.
I forwarded pieces of that email to linux-cirrus@ to let people here
know where things stood, and to let them know that I was still planning
to continue to work on ep93xx peripheral support for the upstream kernel.
When doing so, I took great care not to mention the name of the person
at Cirrus that had sent me that email -- if you re-check the email I
sent to linux-cirrus@ at the time, you'll see that the name "Brian
Austin" isn't mentioned anywhere.
Even though Cirrus has undertaken several attempts to destroy my
reputation by saying and suggesting things that are simply not true,
I hold no ill will against them, and I would still love to see some
form of cooperation between them and the upstream Linux kernel
developers. (Which doesn't have to be me or involve me, if they so
desire.)
> It is instances like that one that keep alot of chip companies from
> working with kernel maintainers.
I myself have worked with Intel engineers, Marvell engineers and
Freescale engineers on upstream Linux support for their ARM SoCs.
(For which two out of those three vendors didn't pay me a dime, BTW,
before we hear that argument again.) (One did pay me, but the work I
did on that was only a small part of a larger set of work I did for
them, most of which not Linux kernel related at all, and besides, I
wonder why I feel the need to defend myself for this.)
I have not personally worked with Atmel, Texas Instruments or Samsung,
but there is support for their ARM SoCs in the upstream Linux kernel
as well.
I don't think Linux kernel maintainers are perfect, but it's certainly
not all so bad. If Cirrus chooses not to support their ARM SoCs in the
upstream Linux kernel tree just because they've butted heads with
someone, then that is of course their right.
Other related posts:
